Non–Intrusive Appliance Load Disaggregation in Smart Homes Using Hybrid Constrained Particle Swarm Optimization and Factorial Hidden Markov Model

Nowadays, the prediction of the load performances in the smart systems is necessary to generate the minimum energy. In a smart home, there are various appliances that each of them has different behavior. These differences defined as appliance states. In this paper, an effective hybrid method is proposed for load disaggregation of appliances. Factorial Hidden Markov Model (FHMM) with high accuracy is used for appliances states modeling. In this model, the present state of each appliance is available, and then the defined allowable states for the next instant are provided. For optimal estimation of states, the Particle Swarm Optimization (PSO) algorithm is employed. Furthermore, three constraints are applied in PSO to modify the states matrix; first, every appliance must has one state at any instant; second, considering of the appliances that always is active; and last, using of FHMM for load models production. In the last constraint, by using FHMM, counts of the estimated databases as well as the calculation time are remarkably reduced. In order to show the effectiveness of the proposed method, speed and accuracy of the responses for practical data of six smart homes are compared with other methods.